Understanding and Estimating the Adaptability of Domain-Invariant Representations
When the test distribution differs from the training distribution, machine learning models can perform poorly and wrongly overestimate their performance. In this work, we aim to better estimate the model’s performance under distribution shift, without supervision. To do so, we use a set of domain-in...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139150 |
| _version_ | 1826190843186249728 |
|---|---|
| author | Chuang, Ching-Yao |
| author2 | Jegelka, Stefanie |
| author_facet | Jegelka, Stefanie Chuang, Ching-Yao |
| author_sort | Chuang, Ching-Yao |
| collection | MIT |
| description | When the test distribution differs from the training distribution, machine learning models can perform poorly and wrongly overestimate their performance. In this work, we aim to better estimate the model’s performance under distribution shift, without supervision. To do so, we use a set of domain-invariant predictors as a proxy for the unknown, true target labels, where the error of this estimation is bounded by the target risk of the proxy model. Therefore, we study the generalization of domain-invariant representations and show that the complexity of the latent representation has a significant influence on the target risk. Empirically, our estimation approach can self-tune to find the optimal model complexity and the resulting models achieve good target generalization, and estimate target error of other models well. Applications of our results include model selection, deciding early stopping, error detection, and predicting the adaptability of a model between domains. |
| first_indexed | 2024-09-23T08:46:10Z |
| format | Thesis |
| id | mit-1721.1/139150 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T08:46:10Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1391502022-01-15T04:05:02Z Understanding and Estimating the Adaptability of Domain-Invariant Representations Chuang, Ching-Yao Jegelka, Stefanie Torralba, Antonio Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science When the test distribution differs from the training distribution, machine learning models can perform poorly and wrongly overestimate their performance. In this work, we aim to better estimate the model’s performance under distribution shift, without supervision. To do so, we use a set of domain-invariant predictors as a proxy for the unknown, true target labels, where the error of this estimation is bounded by the target risk of the proxy model. Therefore, we study the generalization of domain-invariant representations and show that the complexity of the latent representation has a significant influence on the target risk. Empirically, our estimation approach can self-tune to find the optimal model complexity and the resulting models achieve good target generalization, and estimate target error of other models well. Applications of our results include model selection, deciding early stopping, error detection, and predicting the adaptability of a model between domains. S.M. 2022-01-14T14:52:54Z 2022-01-14T14:52:54Z 2021-06 2021-06-24T19:20:15.437Z Thesis https://hdl.handle.net/1721.1/139150 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Chuang, Ching-Yao Understanding and Estimating the Adaptability of Domain-Invariant Representations |
| title | Understanding and Estimating the Adaptability of Domain-Invariant Representations |
| title_full | Understanding and Estimating the Adaptability of Domain-Invariant Representations |
| title_fullStr | Understanding and Estimating the Adaptability of Domain-Invariant Representations |
| title_full_unstemmed | Understanding and Estimating the Adaptability of Domain-Invariant Representations |
| title_short | Understanding and Estimating the Adaptability of Domain-Invariant Representations |
| title_sort | understanding and estimating the adaptability of domain invariant representations |
| url | https://hdl.handle.net/1721.1/139150 |
| work_keys_str_mv | AT chuangchingyao understandingandestimatingtheadaptabilityofdomaininvariantrepresentations |